Interview with Joel Lüthi and Virginie Uhlmann

At the end of 2023, the BioVisionCenter, a new hub for bioimage analysis, was launched with a kick-off symposium in Zurich. We caught up with Joel Lüthi, Head of Research and Development, and Virginie Uhlmann, the center Director, to find out more about the BioVisionCenter, how they each got into bioimage analysis and their advice on leadership and mentorship. The interview was conducted just before Virginie moved to the BioVisionCenter in February.

What first inspired you to become a scientist?

JL: For me, it was my great high school biology teacher René Wunderlin who was able to trigger a fascination and initial interest in the sciences. After that there many things, including reading science fiction books, which gave me a fascination for all the things that could be possible!

VU: I always find this a difficult question to answer. I don’t think I ever consciously thought, ‘I want to become a scientist’, because I didn’t quite know this was a job. I think I ended up doing science because I was fortunate to have lots of people in my path who knew that could be a job for me and who encouraged me to try.

Can you describe your career path since high school?

JL: I can start because I think my version is simpler than Virginie’s! After high school I decided to study biology and bioinformatics at the University of Zurich. After my undergraduate studies, I also did a PhD here at the University of Zurich in the lab of Lucas Pelkmans with a focus on image analysis in stem cells during cardiac differentiation. One key thing I had to figure out was how to handle terabytes of 3D images. This has kept me busy for a while because after my PhD, I moved to the Friedrich Miescher Institute for Biomedical Research (FMI) and worked in the lab of Prisca Liberali as an image analysis expert. During this time, I started to build up the Fractal platform that we’re now continuing to develop at the BioVisionCenter.

VU: To describe my career path I have to go back to the importance of mentors and supportive people on my path! During high school, I turned out to be pretty good at maths and my maths teacher was very enthusiastic in pushing me to do more. He told me that I should try to go to university and to apply to the mathematics department, but I was really terrified of this! Although I wanted to do biology, he managed to convince me that there would be a good mixture of both if I went to EPFL (École Polytechnique Fédérale de Lausanne), which, back then, had just opened their life science faculty. So, I went there and that was a lot of fun. It was during my bachelor’s that I had my first experience of real wet-lab biology and I realised that I really didn’t want to do this type of work! From that point on, I switched to much more maths and computer science. I continued at EPFL to do my PhD with Michael Unser, who works on theoretical aspects of image processing. I moved very much into the theoretical part, but one thing that I always liked was that I always had this connection with biology through the images. I really like the textbook parts of biology, reading about biology, hearing about biology, reading papers, I love this, but I just don’t want to be the person doing the experiments! At that time, bioimage analysis was kind of a nascent field. We were a small, very enthusiastic community, and we realised that something interesting was starting. I had the chance to connect with Anne Carpenter at the Broad Institute and Fred Hamprecht in Heidelberg, and to spend some time with both of them just before and during my PhD. Towards the end of my PhD, it was time to find a job and I was asking myself what I should do. My first thought was to move to industry, but this was when another great mentor came in. I received an email, out of the blue, from Ewan Birney, the Director of EMBL-EBI, saying ‘hello, I heard that you are doing image analysis, we have openings for group leader positions at EBI and you should apply’. I thought, no, no, no, that sounds very scary, but again, I was fortunate to have someone pushing me to try. It turned out to be a great opportunity and I started my lab there in 2018. Since then, I got involved into a couple of other projects, including leading an initiative at EMBL to promote theoretical approaches to biology. I got really excited about leading community initiatives as well as doing research, so I took on the role of Deputy Head of Research at EMBL-EBI. I then heard about exciting things around bioimage analysis happening in Zurich, and that is what brought me to the BioVisionCenter.

Do you find that having a background in biology helps when you’re communicating with your collaborators?

VU: Yes, 100%. To me, one of the really exciting things about bioimage analysis is that it is a genuinely interdisciplinary field. You can’t be doing good work if you don’t understand both the computational and biology side. It took a long time for me to stop feeling insecure about the fact that I’m not a true computer scientist and also not a true biologist. Seeing the field of bioimage analysis be gradually more recognized was also really precious. The moment where I completely understood and captured the value of my interdisciplinary knowledge was the first time that I supervised a PhD student from computer science: I was really excited by the possibilities of what he would be able to tackle but when we attended a biology talk and he asked me “what’s the endoplasmic reticulum?” as he didn’t have any basic biological training, I realized I had badly underappreciated the value of my biological knowledge!

JL: I can really identify with that. I think bioimage analysis really needs to be at this interface and it was very rewarding for me to find this community. Although I studied biology at university, I was very tempted to study computer science. I included some computer science in my training by doing some bioinformatics and I knew that I wanted to find a place to bring the two worlds of biology and computer science together. I think bioimage analysis is that place; we’re not pure biologists (although it is difficult to say what a pure biologist is because the field is so broad) and we are not pure computer scientists, but we see both sides. I feel like I act as a translator, explaining why something is important, or why it is hard and why it is a relevant problem to solve.

Congratulations on the launch of the BioVisionCenter. Can you tell us about the aims of the centre?

VU:  The BioVisionCenter is a new structure that is being created as a joint initiative between the University of Zurich and the Friedrich Miescher Institute for Biomedical Research. The goal of the BioVisionCenter is to develop infrastructure, materials, and resources to make bioimage analysis, at scale, accessible to all. These are the two important aspects to this, ‘at scale’ and ‘accessible’. These aspects are essential, in my opinion, because of how the field is evolving. For example, five years ago, the bottleneck for people wanting to analyse their bioimage was method development. We needed methods to do things that we couldn’t do and, while this need still exists, we have seen impressive developments of lots of elements of analysis. The integration of machine learning and improved computational capabilities, among other developments, bring us to a point in time where we have plenty of good methods. One key bottleneck now is the need to apply these methods on very large datasets and while managing this is basically an engineering problem, it is not trivial to solve. Secondly, accessing these methods and applying them on a new dataset is also not obvious. These two key problems are that we want to be able to help with. One of the core parts of what we are doing is to develop a platform, called Fractal, which I’ll leave to Joel to talk about in more detail as he is the lead developer! Overall, the platform should provide users with a way to streamline the whole process from converting data that comes out of a microscope into OME-Zarr, the file format that has been accepted as a standard by the community, to then applying different types of processing tasks onto these images and finally sending the results into a viewer for inspection or QC and further analysis. What we are really focusing on is providing the underlying structure that can make this process smooth and straightforward. It’s important to us that whatever we do is open source, so that everyone can contribute and that it is useful for the community. This is our big idea, and on top of that, of course, we’ll have training activities and be available to help the community get their methods into Fractal or deploy it for their analysis.

Joel, can you tell us more about the Fractal platform?

JL: As Virginie explained, the goal of the platform is to tackle the two big problems of scaling image analysis and making it accessible. The two are often in competition and, for us, the challenge is how could we get to a place where we can do both. Fractal is our attempt of building an open-source platform to do this, while being aware that we can’t be the ones that solve everything and that we need to work in this great community of open-source tools and integrate with what is already there. Fractal is centered around processing the next generation file format, OME-Zarr, and looking at how we can process this type of data at scale. If we have standardised data, we can build modular processing tools, which are interoperable because we know how the data is structured. Additionally, we build a web platform around this to make it easy for users to structure their workflows and to submit big datasets to a cluster without being exposed to the computational deep side of interacting with the command line and other similar interfaces. The whole Fractal framework is deployed in a federated fashion: Instead of us hosting a central server, Fractal servers are run where someone stores their data and has their compute capacity available.

How have you found the first few months of setting up the centre? And Joel, how have you found stepping into your first leadership role?

VU: So, I guess I can hand over to Joel for this, because I’m actually not working at the centre yet. I’m not even in Zurich at the moment, so Joel has been the one person on-site, running around and doing many, many things for the past year!

JL:

Well, much of this is new to me. It’s a super interesting learning experience though. My main takeaway is that getting things started is both harder than you’d expect, but also more rewarding than you might initially think. There are so many things to figure out, for example how we structure things and how we fit in the larger context of bioimage analysis. But then I come back to what Virginie said in the beginning about the bioimage analysis community just being a very positive community. It’s always rewarding to have community meetings, conferences, or hackathons and interact with people, to see why we want to do this and why it’s worth investing the time in what we are trying to build.

Do you have any general advice for someone moving into a leadership position?

VU:  I think my advice was in Joel’s previous answer: I’ve learnt that the most important characteristics in a leadership role are to listen and to be patient. I think that it is important to take the time to listen and understand how things are done, but also read the current mood of the environment. Then, one can take this on and actually build stuff. And, of course, being very patient is essential because building things take time – everything takes time!

Collaboration is really a key part of your work. What’s your best advice for getting the most out of collaborations?

JL: For me, the big one is that collaborations are all about people. While we’ve come a long way with being able to interact virtually – having Zoom and webinars etc – in a lot of collaborations, I find nothing beats getting to know the people and having in-person interactions. I’ve found that this helps me better understand what people want or need and what their motivations are. Understanding this means we can continue the work through virtual interactions or in-person and build something that is useful.

VU: I totally agree with this. To me, one key elements of successful collaboration is to work well with the person on a personal level. Communication is another key element and with each collaborator you will be communicating in a slightly different way. It’s really worth putting in the effort and investing the time to develop good ways to communicate, as this will ensure that the collaboration is successful.

We’ve spoken a little about mentors, do you have key people that have mentored you? Has this affected your own mentorship style?

JL: In terms of mentorship in a professional context, my PhD supervisor, Lucas Pelkmans, gave me a lot of freedom to explore. And even though it wasn’t necessarily in the best short-term interest of my PhD project to do more image analysis and build more tools, giving me some leeway to figure out which parts have been particularly interesting was always super helpful. And I would not be here today without having had this chance to explore. Then at the FMI, I was working in Prisca Liberali’s lab and she gave me her trust to go and build something. Having that trust was very empowering. For mentoring others, I focus on building trust and making sure I understand what the other person wants or needs. Additionally, I aim to be open and communicate effectively.

VU: For me, the people in my life that have been the most influential as mentors are the people that have taken the time to understand what I really value, what I care about, and provide advice based on that. I’m trying to do this in my mentorship as well, but it’s often really hard. Sometimes this means recognising that what may seem optimal in terms of career advice might not be appropriate for that person. The times that my mentors were able to give me advice that exactly matched what I needed, and not necessarily what was would have been the best by a ‘consensus’ metric, had a big impact on my career. It takes real empathy to understand someone and to be able to support them in choices that may not be the choice that the mentor would have personally made.

What are you most excited about in bioimage analysis?

VU: For me, the really exciting bit of bioimage analysis is that we use images! They are the first data type, and still one of the main types, that biologists generated. Yet, in the way that modern digital biology has evolved, a lot of the computational biology and bioinformatics has primarily been done on other types of data that moved to the digital world before images. I think there will be something really fundamental and interesting in bringing together data from different modalities such as sequencing, transcriptomics, proteomics together with imaging data and integrate all of that to look at complex systems at different scales in both time and space. We are closer than we have ever been to being able to do this and if I live to see this, I’ll be happy!

JL: I would agree with that and there’s a lot of exciting things out there! The hype at the moment is around new AI tools. But I am actually most excited about the people in the community. I think the thing that makes bioimage analysis so cool is the open-source community we have, and how many amazing people are around doing different types of analysis. The thing that I see on the horizon that I’m excited about is the movement towards more FAIR data and standardisation; this will empower us to do more as a community as it becomes easier to work together.

Going back to the BioVisionCenter, can you tell us about the symposium kickoff and hackathon events that you have already hosted?

VU: So, the symposium was the kickoff event for the BioVisionCenter and marked the beginning of the initiative. The idea was to bring together people from Switzerland and elsewhere in the world who are involved in big initiatives in image analysis or bioimaging, to brainstorm on what the community needs, to identify what we are collectively excited about, and so on. I think it went extremely well, and again I think it was a nice demonstration of this very good vibe that this bioimaging community has. It was clear from the talks that there are absolutely fantastic things going on in the US, in Europe, and in various places of the world, and all of the people working on these different initiatives are really keen to connect and work out what are the best ways to move forward together as a community. It is clear that there is a mosaic of different initiatives that potentialize each other and can grow together. Seeing this at the symposium was very exciting!

JL: For the hackathon, and actually the symposium as well, we were pleasantly surprised by the amount of interest it generated. When we started planning them, we thought let’s hope the local people want to come. It turns out, the local people wanted to come, European people wanted to come and for the hackathon, we had people from the US and someone from Singapore! People were really interested in talking about where the field is moving. A big focus of the hackathon were discussions about where the next generation file format is moving and how we standardise and work together. It was super cool to get 45 people together at the hackathon to do work on the OME-NGFF specification and governance, on libraries to process these images and working on different ways of running workflows across the three days of the hackathon. I wished that I could have participated in more of the parts of the event! It was a great way to bring the community together in here in Zurich.

What’s next for the BioVisionCenter?

VU: We are working on recruitment to assemble the core team that will work on the Fractal platform, and also contributing to the OME-Zarr specifications. I’m moving in Zurich next week, so hopefully things will accelerate, and we’ll continue growing. We have a website that is still quite rudimentary, but that at least has an event page, so you can find out about our upcoming events once we have more concrete plans.

Finally, can you tell us something about yourself that is unrelated to your science?

JL: I really like to go skiing. Also, you’ll never hear me say no to getting some reading time for myself, but it is always a struggle to find the time!

VU: I have two pet parrots that I spend a lot of time with. It can be quite funny when I take Zoom meetings from home and people wonder where the noise is coming from!

(4 votes, average: 1.00 out of 1)